OPSpec v0.1.0

Open-Plate-Spectrophotometer

An open-source project to build a 96-well spectrophotometer operating at 620nm. Designed from the ground-up focusing on robust electronic design. Funded by the think[box] Student Project Fund for the 2022-2023 academic year.

Test Prototype (v0.0.2)

This project was born out of an appreciation for Cerillo's portable microplate reader and the arrogant desire to open-source it by recreating what they made on my own. I designed a limited test prototype during the summer (I think sometime in July?) before the school year restarted to see what is required to realise the idea of a small form-factor, portable plate reader.

Concept and Schematic

Functional block diagram

The idea is simple: control 96-element grids of LEDs and light sensors to address individual wells in a plate, convert the signal from the light sensor into something machine parsable, then finally do some ✨math✨ to get an optical density value for each well.

LED and detector grids

I decided to not build a full 96-capable prototype. Instead, I opted to test all the relevant units I needed on a smaller scale. The result was a prototype with all 8 rows (A-H) but only two columns.

Independent control of each element is easily achieved by connecting the grid high and low sides to control MOSFETs: Each row was connected to +5V via p-ch FETs (yes i know, they're wired incorrectly) and each column was connected to GND via n-ch FETs. Each FET gate is assigned its own pin on the MCU to independently control individual elements.

A similar idea is used for switching inputs from specific photodiode elements.

Detectors: photodiodes vs phototransistors

The hard bit for a device like this taking light that passes through the sample wells and converting that into a digital signal. Initially, I planned on using phototransistors to transform a change in incident light intensity into a change in current. But these devices are pretty imprecise and do not have the best linearity. To make phototransistors work, I would have needed to do some really careful biasing -- something I did not think I needed.

The alternative was to use photodiodes. These are a bit better in the sense that they would have required a lot fewer supporting components. They also have pretty nice characteristics:

Plot from Analog ADI

These devices are pretty much linear in their response to incident light intensity: A 10x increase in incident intensity results in a 10x increase in reverse current.

Transimpedance amplifiers and ADCs

Converting current going through a photodiode into digital information is a two-step process. This is nothing new; the typical strategy is to use a transimpedance amplifier to convert transform a current signal into a voltage signal followed by an ADC that can convert the voltage signal into a digital value:

TIA diagram from wikimedia commons.

The transfer function is also pretty simple for a TIA: $$ V_{out} = -I_{in}R_f $$

This is implemented as:

The principle is the same, just compressed a little. The current source is replaced by the photodiode, mulit-channel opamps are used to save space, and $V_{out}$ passes through a standard Schottky diode before being connected to the ADC inputs.

I used an 8ch ADC to independently connect each row for pseudo-switching of the detector outputs. High-side p-ch FET switches control V_cc for the TIAs.

The 12-bit ADC converts output voltage into a digital signal referenced to a stable 5V signal generated by a voltage reference amplified by a low-noise voltage follower:

The MCU, an arduino nano, communicates with the ADC over I²C.